计算机应用与软件
計算機應用與軟件
계산궤응용여연건
COMPUTER APPLICATIONS AND SOFTWARE
2014年
3期
146-150
,共5页
模式识别%核距离%稀疏表示%线性组合%交通标识识别
模式識彆%覈距離%稀疏錶示%線性組閤%交通標識識彆
모식식별%핵거리%희소표시%선성조합%교통표식식별
Pattern recognition%Kernel distance%Sparse representation%Linear combination%Traffic sign recognition
提出一种新的基于核距离的稀疏表示识别方法,方法分为两个阶段:首先计算测试样本与训练样本之间的核距离,并挑选出 M近邻;然后将测试样本用挑选的 M近邻进行线性表示,根据每类训练样本的贡献进行分类。在德国交通标识数据库上的对比实验表明,该方法的识别率优于传统的 PCA、LDA 和 OMP 方法,识别率达到94.2%。
提齣一種新的基于覈距離的稀疏錶示識彆方法,方法分為兩箇階段:首先計算測試樣本與訓練樣本之間的覈距離,併挑選齣 M近鄰;然後將測試樣本用挑選的 M近鄰進行線性錶示,根據每類訓練樣本的貢獻進行分類。在德國交通標識數據庫上的對比實驗錶明,該方法的識彆率優于傳統的 PCA、LDA 和 OMP 方法,識彆率達到94.2%。
제출일충신적기우핵거리적희소표시식별방법,방법분위량개계단:수선계산측시양본여훈련양본지간적핵거리,병도선출 M근린;연후장측시양본용도선적 M근린진행선성표시,근거매류훈련양본적공헌진행분류。재덕국교통표식수거고상적대비실험표명,해방법적식별솔우우전통적 PCA、LDA 화 OMP 방법,식별솔체도94.2%。
We present a recognition method in sparse representation which is based on kernel distance.The method can be divided into two phases.First,it calculates the kernel distance between the test sample and the training samples and selects the M nearest neighbours. Then it uses the determined Mnearest neighbours as the linear representation of the test sample,and classifies according to the contribution of each class of training sample.Comparative experiments made on German traffic sign database show that the method is better than traditional methods such as PCA,LDA and OMP in terms of recognition rate,the recognition rate reaches 94.2%.